The Results of Experiments on

Seated Beam Connections

CHARLES W. ROEDER AND RONALD H. DAILEY

Seated beam connections have been used for many years. TABLE 1
They have historically been designed by tabular methods,1 LRFD Web Crippling Capacity—N=3½ in.
and this tradition was recently continued in the new LRFD Web t f/t w = 1 t f/t w = 1.5 t f/t w = 2
Manual.2 The derivation of these design tables are discussed Thick- Beam Depth Beam Depth Beam Depth
elsewhere3 and will not be repeated here. ness 10 12 16 10 12 16 10 12 16
It should be noted that the tables have historically 3/16 22 20 18 17 16 15
considered local yielding of the beam web under edge 4/16 39 36 32 33 30 27 29 27 25
5/16 61 56 49 51 47 43 46 43 40
loading, the strength of bolts or welds, the yielding capacity
6/16 88 81 71 73 68 62 66 62 57
of the seat angle and the normal detailing dimensions. 7/16 120 110 97 100 93 84 89 84 78
The new AISC LFRD Specification includes the new Resistance factor = .75—Load inkips for A36 steel
equation (K1-5) which checks the beam web for crippling or
buckling. In this paper, the term buckling is defined to mean The stiffened seat prevented yielding of the seat angle and
a vertical crimping or crippling of the web due to the assured that failure occurred in the beam rather than in the
application of a concentrated load on the flange. This is a connection. The top angle was L4 × 4 × ¼ and was attached
stability mode of failure rather than a web yielding failure as to the top flange or optional web location as indicated in
presently considered in ASD design (8th Edition Manual). Table 2. These two locations are required by the AISC design
Unlike a web yielding formula, it includes the depth and the procedure. The top or side angle is required to support the
flange thickness of the beam as variables. Table 1 beam, but the design calculations3 do not explicitly consider
summarizes LRFD beam web capacities according to Eq. it. To evaluate this effect, a third support configuration (Fig.
(K1-5) and a 3-½ in. bearing length. It can be seen that there 2) was used for each of the six sections. This third
are cases where the tabulated unstiffened seat capacities configuration employed 2 angles which were bolted to the
exceed the beam web buckling capacity. Therefore, AISC column but not to the beam. The angles guaranteed the
funded a research program at the University of Washington stability of the beam web, but they did not contribute to the
to study these connections. This paper summarizes some of capacity of the connection seat. Therefore, this configuration
the results of this experimental program. was the most realistic in evaluating the capacity of the
connection independent of any beam web attachments.
EXPERIMENTAL PROGRAM Four additional tests were performed with unstiffened
Six wide flange beams of grade 50 steel were tested in seat angles. The beam was a W14 × 26 or a W14 × 22 for
both stiffened and unstiffened seat connections. They were these tests, and the seat angle was ¾-in. thick. Three strain
the W8 × 10, W12 × 14, W12 × 22, W12 × 26, W14 × 22 gages were attached to the seat angle and monitored during
and W14 × 26. These sections were chosen because they the test. These gages were placed on the inside of the leg of
have thin webs and their predicted load capacity, with ½ in. the angle, which was bolted to the column, near the toe of the
bearing would be governed by web buckling, Eq. (K1-5). The fillet. They were spaced across the width of the angle so that
tests are summarized in Table 2, and the test setup is the distribution of strain could be measured. Three of the four
illustrated in Fig. 1. Each of the six sections were tested with flexible beam tests were made with the standard ½-in.
three different top angle arrangements and a stiffened seat. setback between the beam end and column face. The top
angle was located either on the top flange (standard), or on
the web (optional location) or an unbolted double angle detail
Charles W. Roeder is professor of civil engineering, University of
was used as shown in Fig. 2. The fourth unstiffened seat test
Washington, Seattle, Washington.
Ronald H. Dailey is a former graduate student, Department of used a 1-in. setback between the beam and the column face.
Civil Engineering, University of Washington and currently a partner This fourth test illustrated the effect of fabrication error or
of Loftus and Dailey, Consulting Engineers, Fairbanks, Alaska. inadvertent eccentricity on the connection detail. The bolts

90 ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION

© 2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without the written permission of the publisher.
 Figure 1

were all ¾ in. A325 with calibrated washers. Lateral support
was provided to the beam in the region of maximum moment
with a light frame as shown in Fig. 3. The loads were applied
with a 2.3 million pound Baldwin hydraulic test machine, and
a bearing stiffener was used to prevent damage to the beam at
the load point. An elastomeric pad was used to distribute the
load and prevent local damage. The concentrated load was
applied near the quarter point of the eight foot span. This
loading was needed to develop the required end reaction of
the beam without yielding the beam in bending.
Deflections were measured at three locations in the beam
span with linear voltage displacement transducers (LVDTs)
as shown in Fig. 1. The data were recorded with a Hewlett-
Packard computerized data acquisition system. The test spec-

TABLE 2
Experimental Program
Beam Seat Top Angle Ultimate
Size Detail Detail Reaction
W8×10 Stiffened Top Angle 18. A
W8×10 Stiffened Optional Web 18.3 A
W8×10 Stiffened Unbolted Web 18. A
W12×14 Stiffened Top Angle 36.8 A
W12×14 Stiffened Optional Web 36.3 A
W12×14 Stiffened Unbolted Web 27.6 A
W12×22 Stiffened Top Angle 61.4 A
W12×22 Stiffened Optional Web 60.3 A
W12×22 Stiffened Unbolted Web 60.8 A
W12×26 Stiffened Top Angle 64.5 B
W12×26 Stiffened Optional Web 68. C
W12×26 Stiffened Unbolted Web 57. C
W14×22 Stiffened Top Angle 58.5 C
W14×22 Stiffened Optional Web 57. C
W14×22 Stiffened Unbolted Web 54. C
W14×26 Stiffened Top Angle 60.9 C
W14×26 Stiffened Optional Web 58.5 C
W14×26 Stiffened Unbolted Web 45. C
W14×26 ¾L Top Angle 47. D
W14×26 ¾L Optional Web 47. D
W14×22 ¾L Unbolted Web 38. D
W14×26 ¾L Top Angle 33. D
A—Beam yielded under load point
B—Local flange buckling of the beam
C—Beam web buckled at seat
D—Angle seat bent and could not accept more load.

THIRD QUARTER / 1989 91

© 2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without the written permission of the publisher.
 imens were all whitewashed and all yielding and buckling
was carefully observed and photographed. The configurations
were always tested in a given sequence. First, the beam was
tested with the top angle, and the web and unbolted angles
were tested sequentially. The first two tests were stopped
before excessive damage was done to the steel. This
permitted the use of the same steel section for three tests.
Obviously, a slightly larger load capacity could have been
attained if more damage had been permitted in the first two
tests, but examination of the force-deflection behavior
indicates that this effect is very small. More detailed data on
the test procedures and test results are given elsewhere.
EXPERIMENTAL RESULTS
The experiments were performed and some of the more
general results are noted in Table 2, but several observations
require more elaborate description.
First, it should be noted that comparison of the reaction
capacity obtained for the double angle support system shown
in Fig. 2 with the reaction capacity obtained for identical seat
connections with the top flange angle or the optional web
location angle, illustrates the percentage of the load carried
by the top angle. This effect is further illustrated in Figs. 4
and 5 for typical stiffened and unstiffened seat connections.
The top angle carried between 8% and 35% of the end
reaction for connections which failed through web crippling.
Further, the top angle or optional web angle may stiffen the
seat somewhat in the elastic range and, therefore, delay
inelastic action. The reaction carried by the top angle is
important to the connection behavior even though it is
ignored in the design model used to develop the connection
load tables. However, not enough data is available at the
present time to warrant a less conservative model than that
presently used.
The second observation is that web buckling is an
important mode of failure for the case when beams with thin
webs bear on stiffened seats. For this reason, the AISC LR-
Figure 2
92 ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION
© 2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without the written permission of the publisher.
FD Manual of Steel Construction design procedure includes a
method for calculating the bearing length to prevent both
beam web yielding and buckling. To facilitate this
calculation, AISC has tabulated constants in the Load Tables
for Uniform Load Constants for beams. Unless this tabulated
bearing length is considered, web buckling may be the limit
state as illustrated in Table 2 and the photograph of Fig. 6.
The buckling was typically accompanied by yielding of the
steel and resulted in large permanent deformations. The
deformation probably contributed to the ability of the top
angle to carry load. While web buckling is an important limit
state, it requires a relatively unusual loading condition to
achieve it. The experiments were performed with a
concentrated load at approximately the quarter point of an 8
ft. span. This span is quite short for a 12 or 14 in. wide
flange beam. Further, the quarter point loading is relatively
uncommon and despite these unusual conditions the W8 × 10,
W12 × 14 and W12 × 22 still failed by flexural yielding
before web buckling could occur. Web buckling is unlikely to
occur with rolled wide flange beams under uniformly
distributed loading unless the span is very short compared to
the beam depth. In any case, beam web buckling can be
avoided by following AISC Specification Eq. (K1-5) and
using the procedure outlined in the Manual.
A third major observation is that unstiffened seat angle
connections can adequately yield without fracture. The
connections are capable of rotating, without distress, far
beyond that required to accomodate the end rotation of the
beam it is connecting. The strain gages indicate that yielding
initiates near the toe of the fillet in the center of the leg of the
angle which is bolted to the column. Yielding propagated
along the length of the angle as indicated in Fig. 7, and again
the top angle reduces the load on the seat angle as illustrated
in Fig. 5. The experiments suggest that yielding of the seat
angle occurs near the toe of the fillet on the leg which is
bolted to the column; this measurement is supported by the
observed plastic rotation at the location in the photo of Fig.
8. It is interesting to note that the measured strains never ex-
Figure 3
 TABLE 3
Comparison Of Experiments With Design Values
Beam Seat Top Angle Ultimate ASD DESIGN LRFD Design
Size Detail Detail Reaction Yield Ratio Yield Eq. (K1-5)
W12×26 Stiffened Top Angle 64.5 38 1.7 65 29
W12×26 Stiffened Optional Web 68. 38 1.8 65 29
W12×26 Stiffened Unbolted Web 57. 38 1.5 65 29
W14×22 Stiffened Top Angle 58.5 38 1.5 65 31
W14×22 Stiffened Optional Web 57. 38 1.5 65 31
W14×22 Stiffened Unbolted Web 54. 38 1.4 65 31
W14×26 Stiffened Top Angle 60.9 42 1.4 75 34
W14×26 Stiffened Optional Web 58.5 42 1.4 75 34
W14×26 Stiffened Unbolted Web 45. 42 1.1 75 34
W14×26 ¾L Top Angle 47. 27 1.7 49 34
W14×26 ¾L Optional Web 47. 27 1.7 49 34
W14×22 ¾L Unbolted Web 38. 27 1.1 49 31

ceeded 0.005 in./in. on the outer fibers, far less than the onset are not included in Table 3 because these tests did not result
of strain hardening. in a beam web failure.

VALIDITY OF SPECIFICATION PROVISIONS
The experimental results are compared to the ASD
provisions (1978 Specification) and the LRFD provisions in
Table 3. The LRFD design limits include the appropriate
resistance factors. Under the ASD design rules, only beam
web yielding is considered. The LRFD rules consider both
web yielding as well as buckling calculated by Eq. (K1-5),
using only a 3-½ in. bearing. It is clear that, for both the
stiffened and unstiffened case, web buckling should be
considered. It is also interesting to note that, in the ASD
case, the two cases where the ratio of Ultimate Reaction to
the Web Yielding Load was only 1.1 were cases not
permitted by the ASD Specification (no top or side angle).
The test results for the W8 × 10, W12 × 14 and W12 × 22
Stiffened Seats
Comparison of the stiffened seat test results with the
LRFD factored resistance shows that the LRFD beam web
buckling rule, Eq. (K1-5), is conservative but can not be
ignored. The specimens with top flange angles or optional
web angles had measured strengths more than 72% greater
than predicted by the LRFD rules, including web buckling. If
web buckling was not considered in the design calculations
only one test in six achieved the computed strength. These
observations suggest that a check for web buckling is
essential but it is also apparent that LRFD Eq. (K1-5) is
probably too conservative. This equation was analyzed in
greater detail and compared to 69 other experiments in other
papers.4,6,7 There was great scatter in the experimental data
and, when the statistical concepts of LRFD are employed,
this scatter results in a small (0.75) resistance factor. The L-

Figure 4 Figure 5

THIRD QUARTER / 1989 93

© 2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without the written permission of the publisher.
 RFD resistance for web buckling and yielding; and the
smaller of the two values divided into the measured strength
for the 69 experiments.5 The figure clearly shows that
strengths much larger than the design values were often
achieved. Most of the experiments were performed more than
40 years ago. It would appear reasonable to ask if a new
series of tests could reduce the scatter through good
experimental control and permit a considerable increase in
the assumed resistance and, in turn, the resistance factor.
The third specimen in each group of three stiffened seat
tests had unbolted double web angles as shown Fig. 2. This
test configuration is more nearly consistent with the design
model used to generate the tables for unstiffened seats.
However, as has been pointed out, it is not permitted by
either the AISC ASD or LRFD Specification. The double
angles are not able to carry any part of the reaction, and as a
result the reaction is reduced by 7% to 27% over the seat
when provided with a bolted top or side angle. These tests
illustrate the importance of a securely attached top angle
which is of appropriate size for the connection. This third
detail is still very conservative when compared to the web
buckling equation. This observation supports the earlier
observation that the resistance factor of 0.75 may be small
for this application.
Similar observations may be made with respect to the
allowable stress (ASD) design provisions.1 ASD does not
require a web buckling check and, without the top or side
angle, an unconservative design may result for beams with
slender webs. Even with these angles, two tests indicated a
load factor of 1.4. This again suggests that, even though no
failures have been reported, a web buckling provision should
probably be included in the ASD Specification.
Unstiffened Seats
Only three tests are available for the unstiffened seat
connection, since the 1 in. end spacing is not consistant with
Figure 6
94
© 2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without the written permission of the publisher.
the design tables. Beams with a top or side angle developed
measured reactions which were 5% less than the computed
LRFD yield resistance, but they greatly exceeded the
predicted web buckling resistance. The third test in this group
had the unbolted double angle detail shown in Fig. 2. This
detail is consistant with the design assumptions3 used for
unstiffened seats but it is not consistant with the requirement
to provide a top or side angle. The measured strength for the
specimens with the double angle was below acceptable levels
for both the ASD and LRFD design. This again illustrates the
importance of a securely fastened top or side angle. This
angle carries a substantial portion of the end reaction, and the
loss of this capacity appears to result in an unconservative
estimate of the beam web buckling capacity. As noted earlier,
the unstiffened seats were able to deform sufficiently while
carrying the required load. It is concluded, therefore, that the
present or LRFD design procedure for the seat, itself, is
adequate. However, it is possible for the tabulated values to
exceed the buckling capacity of some beams with slender
webs so it is essential that LRFD Eq. (K1-5) be checked.
CONCLUSIONS AND PRACTICAL IMPLICATIONS
It can be concluded from the results of this research
that:
1. The top flange angle (or optional web angle) carries a
substantial portion of the end reaction for both stiffened
and unstiffened seat connections. This makes it essential
that this angle be securely fastened and be of an
appropriate size and strength for the connection.
2. Web buckling is the controlling failure mechanism for
many beams with slender webs and small end bearing
lengths.
3. ASD design provisions do not explicitly include a beam
Figure 7
ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION
 web buckling check and as a result, unconservative
designs could result, particularly if the required top or
side angle were omitted.
4. LRFD provisions have an equation for web buckling,
Eq. (K1-5), but the experiments described in this paper
suggest that this equation with a 0.75 resistance factor
may be overly conservative. However, it must be
emphasized that 69 tests performed 40 years ago and
described in other papers3,6 suggest that this formula
and resistance factor are approximately correct.
5. The three tests of unstiffened seats indicated that the
present design procedure results in seats that have
ample rotation capacity and strength.
Practical Implications
The previous discussion suggests that the existing design
rules for unstiffened seat connections are somewhat
irrational. They employ a highly idealized and probably
unrealistic distribution of forces and they sometimes ignore
the critical mode of failure. This conclusion is unavoidable
when the experimental evidence is considered. However, it
important that one not lose sight of several important
practical aspects. These connections have been used for many
years and no problems have been noted and one must
recognize why this is so. First, the buckling failure mode
described in this paper requires a high shear force and a
beam with a slender web. This would be unlikely to occur
with rolled shapes loaded uniformly. The limit states would
be more likely be a three hinge mechanism in these common
applications. Secondly, and very important, the top or side
web angle carries a significant portion of the end shear, and,
if it is properly bolted and not too thin, it will help solve
many potential problems. Finally, web buckling is most likely
to occur with stiffened seats or unstiffened seats with thick
angles and when the requirement for a top or side angle is
violated. The unlikely combination of all these factors
explains why these connections have had such a good
performance record.
RECOMMENDATIONS
More research is needed to improve the design of seated
connection through:
1. Developing models for predicting the effect of the
top or side angle on the connection strength.
2. Better estimates of the web buckling capacity and the
LRFD resistance factor. There is a great deal of
scatter in the existing data3 and this data was
acquired more than 40 years ago. A well controlled
experimental program may well reduce this scatter
and provide more realistic strength estimates.
3. Development of better models for predicting the
distribution of forces in the connection material.
ACKNOWLEDGMENTS
The experiments described in this paper were performed
with funds provided by the American Institute of Steel
Construction.

REFERENCES
1. Manual of Steel Construction, 8th ed., Chicago: AISC,
1980.
2. Steel Research Needs for Buildings, NBS Special
Publication 693, U.S. Dept. of Commerce, Washington
D.C., 1985.
3. Load and Resistance Factor Design Specification for
Structural Steel Buildings, Chicago: AISC, 1986.
4. Elgaaly, M., "Web Design Under Compressive Edge
Loads," AISC Engineering Journal, Vol. 20, No. 4,
Chicago, 1983, pp. 153-171.
5. Roberts, T. M., "Slender Plate Girders Subjected to Edge
Loading," Proceedings Institution of Civil Engineers,
Part 2, Vol. 71, September 1981, pp. 805-819.
6. Ketchum, M. S., and J. O. Draffin, "Strength of Light I-
Beams," University of Illinois Engineering
Experimentation Station, Bulletin No. 24, February
1932.
7. Lyse, I., and J. H. Godfrey, "Web Buckling in Steel
Figure 8 Beams," Transactions, New York: ASCE, 1935.

THIRD QUARTER / 1989 95

© 2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without the written permission of the publisher.